A Study on the Flexibility of Electricity Consumers for the Swedish Context – Modelling, Quantification and Analysis of Notice Time Lars Herre Tommy Kovala Christos Papahristodoulou Lennart Söder KTH Royal Institute of Technology Mälardalen University
“Time is what we want most, but what we use worst.” - William Penn (1644-1718) 8/31/2016 2
Agenda 1. Motivation • Wind Power Forecast Horizon 2. Demand Response Programs • Notice Time 3. Modelling of flexible consumers • Demand Bidding with updates 4. Elasticity Review 5. Elasticity Results 6. Discussion & Conclusion 7. Outlook 8/31/2016 3
1. Motivation 8/31/2016 4
1. Motivation Demand • Renewable Energies • Intermittent Production • Uncertain Forecasts Solar Wind 8/31/2016 8/31/2016 5 5
1. Motivation: Wind Power Integration • Wind power forecast accuracy: • Probability for more accurate forecasts improves Root Mean Square Error 8/31/2016 6
1. Motivation: Wind Power Integration Forecasts 8/31/2016 7
1. Motivation Balancing in different time frames: • Historically: • Hydro power • CCGT turbines • Future: • Hydro • Storage (el., chem., etc..) • Demand Side 8/31/2016 8
2. Demand Response Programs • Market-based • Dispatch-based • Control-based • Time-based 8/31/2016 9
2. Demand Response: Notice Time • Notice Time: time span between notification and delivery (Process Management: Order Lead Time) The stages of a Demand Response Event, adapted from: [Source: SEDC, “Mapping Demand Response in Europe Today,” 2014.] 8/31/2016 10
2. Demand Response: Notice Time • Assume ε to decrease as notice time decreases -1.0 Elasticity ε -0.1 / Notice Time 8/31/2016 11
3. Continuously Updated Demand Bidding 8/31/2016 12
3. Continuously Updated Demand Bidding 8/31/2016 13
3. Modelling • Continuously Updated Demand Bidding: wind decrease 8/31/2016 14
3. Modelling • Continuously Updated Demand Bidding: wind increase 8/31/2016 15
4. Elasticity Review Estimated price elasticity • Exclude non-price related effects Own-price elasticity for instant response • Estimated for the simple setups of Time-Of-Use pricing and the complex setups of Real-Time-Pricing. Time 00-06 06-08 08-10 10-16 16-18 18-20 20-00 Elasticity E1 E2 E3 E4 E5 E6 E7 Time Low Peak Period Low Elasticity E1 E2 E1 8/31/2016 16
4. Elasticity Review Assumptions 1. Different approaches for elasticity estimation give similar results. – Difficult to find projects/articles with similar approaches and models from which elasticity has been estimated. – A general lack of published own-price elasticity estimates. 2. The relation between elasticity estimations from different setups tested in the same project should be similar to other projects. Notice times described in phrases like “Day - ahead” can 3. be associated with a certain number of hours. 8/31/2016 17
4. Elasticity Review In 2012/2013 EU’s Joint Research Center reported from 281 smart grid projects in 30 European countries. Only 65 (23 %) of the projects involved consumer engagement. • Projects have included up to 2000 participating consumers, generally biased towards volunteers with specific interest in this development. • In the 65 consumer engaging projects in Europe two projects included any version of the word ELASTICITY in the project description. – Commercial and industrial consumer 8/31/2016 18
4. Elasticity Review Summary of results from residential US DR-pilots (Shariatzadeh et al., 2015): Maximum peak reductions percentages reached • Avg. for projects including technological assistance: 38 % • Avg. for projects without technological assistance: 27 % • Avg. for Critical Peak Pricing: 34 % • Avg. for Time Of Use pricing: 17 % • Avg. for Real Time Pricing: 16 % There are apparent benefits for certain program setups! 8/31/2016 19
4. Elasticity Review PREVIOUS PROJECTS : A search on Google Scholar for PRICE ELASTICITY ELECTRICITY ”DEMAND RESPONSE” has almost 7000 results • Early studies (1977- 2000) on the “close -to- hourly” price elasticity of households’ electricity consumption seem to have been dominated by Time-Of-Use (TOU) pricing projects. • Lately Critical-Peak-Pricing (CPP) and Real-Time-Pricing (RTP) have taken over the scene. • Concluded: residential consumers have lower own-price elasticity for RTP-programs than for TOU pricing, a result of information costs. (1) • Difficult to compare the results and price elasticity estimates from these different projects 8/31/2016 20
4. Elasticity Review ”Notice time” -dependent elasticity • TOU has a static/indefinite notice time, • CPP and RTP use different notice times depending on DR- program structure. • Emphasis in articles on presenting notice times and description of the standards for the communication with consumers has been low. Objective for this study: • Gather the results from at least 15 DR-projects spread over the three types of pricing for DR-programs, where the notice time is explicitly presented. • Analyze the possible relation between notice time and own- price elasticity. 8/31/2016 21
5. Elasticity Results From California SPP (Faruqui, 2005): 1. Adding assisting technology to DR-program : + 15 % (more elastic) 2. Changing from static to day-ahead notice time: - 34 % (less elastic) 3. Changing from static to day-of notice time: + 18 % (more elastic) Other pilots with multiple DR-programs: • (1) is supported by findings in NJ PSE&G RPP (Faruqui and Sergici, 2010) • (3) is supported by findings in Washington GOPP (Hammerstrom, 2007) 8/31/2016 22
5. Elasticity Results 8/31/2016 23
5. Elasticity results Averages without elasticity estimates from old projects (before 2000): • Avg. for all projects: -0.25 • Avg. for projects including technological assistance: -0.11 • Avg. for projects without technological assistance: -0.20 • Avg. for static notice time: -0.34 • Avg. for day-ahead notice time: -0.03 • Avg. for day-of notice time: -0.11 • Not the hypothesized effect from technology…! 8/31/2016 24
5. Elasticity Results 8/31/2016 25
6. Discussion & Conclusion - Elasticity How can comparability be assured in price elasticity estimation for DR-projects? • Challenges – Hourly or averages for all peak hours – Exogenous causes for demand increase – Different price ratios • Suggestion – Presenting detailed results hourly within the peak hours. 8/31/2016 26
6. Discussion & Conclusion CU-DB model: • Enables consumers to be updated & participate • Consumers in competition with flexible generation in forward markets • Opens the opportunity for both bidding parties to increase their utility in consecutive market places • Risk of deterministic consumers is minimized • Wind power producers may exploit market power by strategically bidding with a lowered wind power forecast 8/31/2016 27
7. Outlook • In CU-DB model: both forecast error magnitude and notice time affect the market equilibrium Static Tariffs Day-Ahead Intra-Day Very High High Medium to Low Forecast Error (>10%) (10-5%) (5-1%) Demand Elasticity Medium Lower Higher • Market with updates can be beneficial for improved integration of renewables. 8/31/2016 28
7. Outlook ? 8/31/2016 29
Thank you Lars Herre Tommy Kovala Lennart Söder Christos Papahristodoulou Mälardalen University KTH Royal Institute of Technology 8/31/2016 30
Recommend
More recommend
